Were there any major challenges that you encountered? If so, what were they and how did you overcome them?
Martin Andersson: Oh, heads were rolling; digital heads. At one point the Tusseladd had his three heads merged into one clump, true to the original design. After some time the director wanted separate necks for the heads. This was a huge change and I had to go all the way back to the base mesh to alter it. Most of the detail I had already put in I could re-project with ZBrush's projection features. That was a life saver! Starting everything over wouldn't really have been an option.
Did the fact that three different effects houses were used pose any special difficulties?
Martin Andersson: Not really. We worked on our trolls while they worked on theirs. The external overall VFX supervisor (Øystein Larsen) and the Post-Production Producer (Marcus Brodersen) were visiting all the studios frequently to keep a total overview of how the material developed. It was very exciting to finally get to see the other studio's trolls because we had very few meetings where all the studios showed what they had been working on. We were all focusing on our own trolls.
From the video that's been seen so far, it looks like much of what we see of the trolls is with night vision and shaky cameras. Did this make your work easier, or more difficult?
Samuel Karlsson: From a visual effect stand point, getting plates shot in cloudy daylight, adding your troll(s) and finally creating the night vision look on the comped shot would for a number of reasons probably be more straightforward than adding the effect the way we did it. We actually used a night vision troll on a night vision plate. Seeing the way the night vision device distorts the footage is quite harsh and also different in each channel, producing aggressive chromatic aberration. Added to that, there is heavy noise (also distorted), plus static noise/dirt. And from what you can see, the footage is not really sharp and crisp, either. All these things make the tracking part a bit trickier and more tedious.
From a compositor's perspective, the difference between working with night vision plates or not is a bit different. There are some things you can't completely copy correctly, like the way shapes and colors interact and bleed into each other when you apply a night vision device to the camera. Meaning, the adjacent pixels to the introduced troll in the plate should have an effect on the troll when converting to night vision and the area around the troll in the plate should be affected as well. So you end up having to mimic the look you believe is correct to the extent you can with the tools available. However, if you instead had the "luxury" of having to add the night vision effect in the end on the comp, on both the plate and CG, you still have to match your shot to a certain look or more probably to an existing night vision plate. And having seen how unusual the way night vision behaves sometimes I wonder if in certain cases it is easier to match a comparatively small area of the shot to your plate rather than the whole shot to other night vision footage.
So to answer your question, from a compositor's point of view, perhaps it's more difficult, and then again maybe not. I wish we would have had the time and budget to investigate both scenarios in depth. But in the end, you naturally have to respect and trust the decisions made for the scenes.